“Solving sizing in fashion e-commerce: How to navigate virtual fitting rooms”

The Sizing Problem

Sizing in fashion has long-rooted problems: a brief history

Historical picture of Harrods department store, seen from the street

Harrods department Store, circa 1905

Source: www.commons.wikipedia.org

The history of fashion sizing is engrained in the shift from bespoke tailoring to mass-produced ready-to-wear clothing. Before the 19th century, clothing was always custom-made, with tailors creating garments based on individual measurements. Standardized sizing began with military uniforms in the 1800s, which required efficiency and consistency. Chest measurements were the foundation for men's sizes, but women's clothing remained custom or pattern-based for decades.

The first major step toward standardizing women's sizes came with the USDA's 1939 study, which collected body measurements from thousands of American women. However, the data was skewed due to a lack of diversity among participants and questionable methodologies. The 1949 National Bureau of Standards (USA) study refined this data, leading to the introduction of the “Commercial Standard” in 1958, which categorized sizes into ranges like "Misses" and "Women." Despite these efforts, sizing struggled to account for diverse body shapes and was often criticized for reinforcing the idealized hourglass figure.

Enter the changing or fitting room

Fashion retail began in small, specialized boutiques in 17th and 18th-century Europe, particularly in Paris. These early shops catered to aristocrats, offering custom-made garments and exclusive access to the latest trends. Boutiques allowed customers to interact directly with shopkeepers and try on items, introducing an early version of the fitting experience.

By the mid-19th century, grand department stores like Le Bon Marché in Paris and Harrods in London expanded on this idea, introducing fitting rooms to accommodate a growing middle class eager for ready-to-wear clothing. Changing rooms became a fixture, offering privacy and enhancing the allure of shopping by helping consumers visualize themselves in their purchases.

These early practices laid the groundwork for modern retail habits, and engrained in the modern consumer the habit of trying on clothes (or shoes) to assess their fit.

Sizing inconsistencies: an endemic challenge

Inconsistent sizing across clothing brands is a widespread issue driven by multiple factors related to industry practices and market strategies. One major reason is the lack of universal sizing standards. Brands often develop their own sizing charts to cater to specific customer demographics, creating variations based on their target audience's preferences and body types. For example, a brand catering to younger, fashion-forward consumers may use smaller fit models, while those targeting older or curvier demographics may adopt more generous sizing. This individuality allows brands to differentiate themselves but leads to significant inconsistencies for shoppers.

Inconsistent sizing is not only a result of differing brand strategies but also stems from the complexities of product development processes. A garment's sizing often starts with a single fit model chosen based on the creative director's vision, which may not represent the brand's broader customer demographic. This top-down approach prioritizes aesthetic ideals over inclusivity, creating patterns and sizing that fit the model but fail to align with the diversity of real body shapes within the target audience.

Furthermore, brands that rely on external manufacturers or suppliers frequently delegate the creation of paper patterns to these third parties. While this practice streamlines production, it introduces deviations from the brand's intended sizing because of differences in interpretation, expertise, and regional size norms used by the supplier. These variations are compounded by the absence of robust quality controls to ensure consistency across different production batches. Consequently, the size labeled on a garment may not match the brand's advertised standards, adding another layer of frustration for consumers trying to navigate already inconsistent sizing.

Historical shifts in sizing standards further complicate the issue. Over decades, societal ideals of body shapes have evolved, influencing how garments are designed and labeled. Practices like vanity sizing, where garments are labeled smaller than their actual size to boost consumer confidence, have exaggerated this problem. Additionally, geographic variations, such as differences between U.S., European, and Asian sizing systems, amplify confusion.

Waist, in inches. Chart about the increase in waistline sizes over the years, from 1958 to 2015Bust, in inches. Chart about the increase in bustline sizes over the years from 1958 to 2015

“Impact of Vality Sizing on Waist and Bust” - Charts created by The Washington Post using statistics released by the American Society of Testing and Materials and created by data visualiser, Max Galka

E-commerce's response to sizing issues: the size chart

The fashion e-commerce size chart emerged as online shopping gained popularity in the late 1990s and early 2000s, becoming a key tool for brands to address the absence of in-person fitting rooms.

Initially, these charts were straightforward, providing a general guide based on assumed standard body measurements. However, as online shopping exploded, brands sought to streamline the digital experience, often adopting a single size chart per gender to cover their entire catalog, regardless of product or fit. This practice, while efficient for retailers, often left consumers confused and frustrated, as it failed to provide the nuanced guidance needed for the wide variety of fits and materials offered by modern apparel. It also overlooks critical variations in design, such as slim-fit versus oversized styles, the stretchiness of fabrics, or seasonal differences like winter coats versus summer dresses. A size guide for a stretchy jersey top might not accurately translate to a rigid denim jacket. By treating all garments as if they conform to a single standard, brands inadvertently create a disconnect between consumer expectations and the actual fit of their products.

Size chart for Men's clothing

Men's Ambiguous Clothing Size Chart

Source: www.evo.com

Another persistent issue with fashion e-commerce size charts is the ambiguity over whether they reflect body measurements or garment dimensions. This lack of clarity can mislead consumers; for instance, a "bust" measurement might refer to the circumference of the wearer's chest or the finished garment, which will fit differently depending on ease and cut. This confusion is compounded by inconsistent terminology, leaving consumers unsure how to interpret the provided numbers or how closely they should adhere to the chart when choosing a size.

E-commerce size charts are further undermined by inconsistencies in product development and manufacturing. As previously discussed, deviations in sizing can occur during the pattern-making process, especially when external suppliers are involved. This results in garments that don't consistently match the size chart. Quality control failures can exacerbate this problem, with variations between production runs leading to discrepancies even within the same size. Consequently, consumers who rely on size charts may still end up with ill-fitting garments, discouraging trust in online sizing tools.

Even in the best-case scenario—where size charts are accurate and transparent—they remain underutilized by consumers. Many people lack access to measuring tapes at home or may find the process of taking their own measurements cumbersome. Even when they attempt it, human error is common, particularly with difficult-to-measure areas like shoulders or bust. For example, a woman measuring her bust may drop the tape (or keep it too tight) as she attempts to measure around her back, or someone measuring their shoulders may misjudge the correct points to include (additionally to not being able to reach them).
This combination of effort and potential for error discourages consumers from engaging with size charts, leaving them to rely on guesswork or familiarity with a brand's sizing, which often proves unreliable.

Other resources for the online fashion shopper: pictures and fit information

When shopping online, the imagery of models wearing the product is one of the most prominent tools available to consumers for gauging size and fit. However, these visuals are unreliable due to the priorities behind their creation. Fashion photoshoots are designed to make garments appear as appealing as possible, focusing on aesthetics rather than providing accurate fit information. Models often wear samples, which may not align perfectly with production sizes, and stylists routinely modify garments to improve their on-camera appearance. For instance, pins or clips are used to adjust the fit of the garment to the model's body, masking potential issues like excess fabric or misaligned proportions. Even when brands include the model's height and weight, these details are often insufficient, as they do not account for variations in body shapes or how the fabric drapes differently across diverse builds, and combined with the highly-altered photoshoots can be flat-out misleading.

Man shirt on a mannequin, clipped on the backWoman blouse on a mannequin, clipped on the back

Clipped clothes on mannequins

Source: blog.mannequinmadness.com

Another challenge lies in the frequent absence of essential product information that could help consumers make informed decisions about sizing. Critical details like fabric composition, stretch, and intended fit (e.g., slim, regular, oversized) are often missing from product descriptions. This gap arises from systemic inefficiencies in the fashion production process. Brands frequently fail to transfer data from product lifecycle management (PLM) or enterprise resource planning (ERP) systems into their e-commerce platforms. Even when this information exists in technical documents, it is rarely made accessible to e-commerce systems (e.g. PIM - Product Information Management). As a result, fit descriptions or fabric details are either omitted or inaccurately guessed by content creators, relying on misleading imagery (see photoshoots discussed above) rather than technical specifications. This disconnect between the product's intended fit—shaped during development by patternmakers and creative directors—and the messaging presented online leaves consumers without reliable guidance, compounding their difficulty in selecting the right size.

The Cost of the Sizing Problem

The canary in the coalmine: Returns

The challenges of selecting the correct size in online fashion shopping significantly contribute to high return rates. Practices like "bracketing"—where consumers order multiple sizes or colors and return the unwanted items—are prevalent. In Europe, online fashion return rates vary between 20% and 40%, depending on the category. Dresses and skirts often see return rates exceeding 50%, primarily due to fit issues, which account for 60-70% of returns in categories such as pants, footwear, and dresses.

Fashion items with the highest online return rates in Europe

Fashion items with highest online return rates in Europe in 2022

Source: www.statista.com

Managing these returns is costly for retailers. Expenses include shipping for both the initial delivery and the return, as well as costs associated with inspecting, repackaging, and potentially repairing or cleaning items for resale. For inventory-sensitive products, returns can lead to missed sales opportunities, especially in popular or hard-to-find sizes. Less than 20% of fashion returns result in an exchange, and items returned late in the season often lose value, necessitating steep discounts. Additionally, some returned goods are unsellable due to damage or wear, adding to waste and costs.

The unit economics of returns are significant. Each return costs retailers an estimated $20-$30 per item, considering shipping, labor, and loss of value. According to 2024 NRF's Consumer Returns in the Retail Industry report, for every $100 of merchandise sold online, approximately $17 is returned, with variations depending on the brand and segment. High return rates exacerbate logistical inefficiencies, as returns must be reintegrated into supply chains.

“For every $100 of merchandise sold online, approximately $17 is returned”

Overall, returns represent a substantial financial burden for fashion retailers. In 2023, the global reverse logistics market, which includes returns, reached $700 billion, with fashion accounting for a significant portion. For individual brands, returns often represent 10-20% of gross sales. For instance, a $1 billion brand could face $100-$200 million in annual return-related costs, significantly impacting profitability, especially for brands offering free returns—a common practice historically designed to attract customers. According to a 2023 study by Coresight Research, in the US alone returns cost the Apparel industry 38B$ per year.

To mitigate these challenges, some retailers are implementing return fees. A 2024 study by Virgin Media O2 Business revealed that 60% of British shoppers would stop patronizing retailers if charged for online returns, with 47% having already been charged, some over £5. Although 42% still expect free returns, the trend of return fees is growing due to rising operational costs. Retailers risk losing especially younger shoppers, important for e-commerce growth, as 58% of 18-24-year-olds commonly buy multiple sizes to return non-fitting items.

A less-obvious cost: Customer Service burden

Customer inquiries about sizing place a substantial burden on fashion's customer service teams. Estimates suggest that sizing-related questions account for 25% to 40% of all customer service inquiries in the fashion sector. Resolving these queries often requires back-and-forth communication, taking an average of 5-10 minutes per interaction, depending on the complexity of the sizing question and the availability of detailed product information. With fashion retailers processing thousands of orders daily, this equates to hundreds of hours of customer service time spent weekly on sizing issues alone. This ends up increasing operational costs and delaying service for other customer needs, potentially costing sales to brands.

Opportunity cost 1: Conversion Rates

Sizing uncertainty significantly impacts consumer purchasing behavior in the fashion industry. Approximately 52% of shoppers hesitate to complete a purchase when unsure about fit, making it a leading cause of cart abandonment.

52% of shoppers hesitate to complete a purchase when unsure about fit

Additionally, a 2023 study found that 58% of consumers prefer purchasing from brands whose sizing they are familiar with, highlighting the importance of consistent sizing in driving sales.

This hesitation is particularly pronounced among first-time buyers, who lack prior knowledge of a brand's sizing standards, thereby affecting a retailer's ability to attract and convert new customers.

The financial implications of missed sales due to sizing uncertainty often surpass the costs associated with handling returns. While processing a return costs fashion retailers approximately $20-$30 per item, the gross margin lost from a failed conversion is substantially higher. For instance, with an average transaction value of $100 and a gross margin of 50%, a single lost sale equates to $50 in lost profit—nearly double the cost of processing a return.

In the luxury segment, this ratio is even more pronounced. Moreover, improving conversion rates by addressing sizing concerns can generate additional revenue, which may offset return-related losses and create economies of scale in logistics.

Opportunity cost 2: Average Order Value

Many consumers are reluctant to purchase multiple items or more expensive products when uncertain about fit. According to Invoca research, 81% of consumers feel more confident making purchases, including high-value ones, when they have access to information that resolves their doubts, such as sizing clarity. According to another 2020 study, 49% of consumers are more likely to add 1-2 items to their basket if they can be confident of purchasing the right size. This lack of confidence in sizing often leads to lower Average Order Values (AOV), as customers opt for fewer or less expensive items to minimize perceived risks.

Moreover, a 2020 McKinsey study indicates that when shoppers encounter uncertainties—such as concerns about choosing the wrong size—they may forgo higher-priced products or stick to brands they already trust for consistent sizing. This behavior highlights a significant missed opportunity for retailers, especially when high-value items yield greater margins. In fact, the hesitation around sizing not only reduces the volume of purchases but also shifts demand away from premium categories, exacerbating the AOV gap.

Intangible costs: Brand Loyalty and poor User Experience

The importance of fit for fashion and the difficulty of finding the right size negatively impacts brand loyalty and customer experience by creating frustration, putting consumers through the painful process of managing returns, and eroding trust. When customers experience inconsistent sizing or are unable to find the right fit, they often resort to "bracketing" — ordering multiple sizes of the same item and returning the ones that do not fit. While this practice allows customers to ensure they get the right fit, the hassle of returning items significantly reduces satisfaction. Studies show that 62% of shoppers consider ease of returns and accurate sizing information key factors in their decision to shop again with a brand. Frequent negative sizing experiences can lead to a loss of repeat customers, as dissatisfied shoppers are more likely to explore competing brands that promise better fit reliability.

Moreover, the negative customer experience associated with sizing issues damages long-term brand loyalty by disproportionately affecting new customers. First-time buyers, who lack familiarity with a brand's sizing standards, are particularly vulnerable to frustration. According to McKinsey,

Nearly half of first-time customers who face fit issues do not return to shop again.

This limits retailers' ability to grow their customer base and tap into new market segments. Poor sizing experiences can also lead to negative word-of-mouth, with dissatisfied customers sharing their experiences online, further tarnishing a brand's reputation. The cascading effect of missed sales, reduced loyalty, and customer churn underscores the critical importance of addressing sizing challenges to maintain a positive customer experience and foster brand allegiance.

The cost for the planet

Beyond the economic aspect, the environmental impact of fashion returns is substantial, contributing significantly to carbon emissions and waste. In the United Kingdom alone, according to Statista, online fashion returns in 2022 generated approximately 750,000 metric tons of CO₂ emissions, equivalent to 175,000 gasoline cars driven for one year. Notably, 350,000 metric tons of these emissions were attributed solely to reverse logistics—the process of transporting returned items back to retailers.

A considerable portion of returned fashion items cannot be resold as new due to damage or wear. Draper's estimates indicate that between 3% and 10% of fashion returns are unsellable, with 50% of these items ending up in landfills.

Kenya's biggest landfill

Kenya's biggest landfill

Source: Caitriona Rogerson, on www.thenationalnews.com

This disposal not only contributes to environmental pollution but also represents a significant waste of resources used in the production of these garments.

The transportation involved in the returns process further exacerbates environmental concerns. The emissions from reverse logistics, combined with the initial shipping, amplify the carbon footprint of each returned item. For instance, the carbon emissions from transporting a returned item can be equivalent to driving a typical car for 10-20 miles.

The Solution: enter the virtual changing room

What is a Virtual Try On?

Part of the reason why a consumer enters a changing room in a physical boutique is to see how a fashion item (e.g. a dress) looks on them - enter the Virtual Try On, or VTO. VTO is an emerging technology in the fashion-tech world, and as such it is still loosely defined. The basic concept is to allow a consumer, typically within e-commerce, the possibility of seeing a garment, shoe, or accessory on a virtual person. The virtual person could be a generic avatar (3D or 2D), or an actual digital twin of the consumer. There are 6 main categories of VTOs currently on the market:

Standalone digital asset
This is the most basic type of VTO, and not everyone would agree it falls within the VTO category. This involves creating a digital twin of the product, typically displayed on the product details page. There are two main methods for generating this type of asset.
  1. The first approach is through Photogrammetry (combination of pictures), or the use of a 3D scanner to create a 3D digital twin of the product. This approach works reasonably well with “hard” goods, like shoes and accessories, but not with apparel. The reason for this is that Photogrammetry captures static images to build a static image, and hence has trouble capturing the complexities of fabric behavior, how it moves, texture and elasticity. Producing these assets typically costs $300 to $700 per product, depending on complexity of the product and quality of the output. Examples of players in this space are London Dynamics and Zakeke, amongst many others.
  2. Virtual hat customization

    Product configuration and customization with Zakeke.

    Source: appsumo.com

  3. The second approach to building a digital asset is through 3D design. This involves generating a digital paper pattern by either importing 2D garment patterns (the same used in a paper pattern for physical manufacturing), or drafting them directly in the software, and then adjusting them (e.g. sewn together). The software then uses advanced physics based algorithms to simulate fabric behavior and customize fabric properties. The result is a highly accurate 3D digital twin, that can be accurate not only in shoes and accessories, but also in apparel. If the software is not used throughout the product development process by the brand, recreating this digital asset typically costs $1,500 to $2,500 per product. Examples of software used to build this type of asset are CLO 3D, Browzwear, Marvelous Designer, Style3D, Optitex,Tukatech and Lectra Modaris.
  4. Comparison between a real-life garment and a digital twin made by CLO.

    Digital twin, created by CLO

    Source: www.clo3d.com

Augmented Reality
Augmented Reality, or AR, aims to display a digital twin of the product on the actual body of the consumer, in the “real” world. In other words, it shows an overlap of the product, for example a shoe, on the body (in this example, the foot) of the actual person using the technology, typically through the camera of a smartphone. Digital assets are typically created through Photogrammetry (as described above), which carries a couple of limitations. Firstly, in terms of product. Once again it tends to work only with shoes and accessories. The second is in terms of accuracy. Since the digital twin is an image without actual physical properties, the body's overall width is only approximated. Getting back to the shoe example, this might result in the foot appearing to stick out of the shoe at certain angles. Players in this space include Reactive Reality, WANNA, Vyking, ZERO10, amongst many others.
Augmented reality try-on on SnapChat

Augmented reality try-on on Snapchat

Source: www.time.com

Virtual Reality
Virtual Reality (VR) VTO aims to display the product in a virtual world - think metaverse. The digital asset is typically displayed on a virtual avatar, shown in a fully customizable environment. The digital asset is recreated as a digital model, typically using technology similar to that of the world of gaming. To be visualized, the user needs specific hardware, such as Meta Quest 3 (formerly Oculus) or the Apple Vision Pro. Because of the barrier generated by the hardware, this approach is typically only used with high-ticket and high-complexity products, such as product customization or made-to-measure.
Garment shopping in a virtual world

Augmented reality shopping

Source: www.obsessar.com

Virtual Mirror
The concept of the Virtual Mirror takes us once again into the physical world. A virtual mirror is an interactive screen, often mounted in a physical retail store, with a camera. The camera reflects the image of a consumer standing in front of the Virtual Mirror, and using AR technology, it allows the consumer to overlay a digital version of a product on the reflected image of the consumer. Due to the limitations of the underlying AR technology, this approach is not well-suited for apparel, which is coincidentally the category where it could be most beneficial.
Shopping assistant helping a customer during a smart mirror shopping experience

Smart mirror shopping experience

Source: www.quytech.com

3D VTO with generic avatar
The concept of the 3D VTO is potentially the most scalable solution. It allows the consumer to see a 3D (or in some cases 2D) digital twin of the garment on screen (e.g. mobile, laptop, tablet, desktop), overlaid on a 3D avatar. The digital asset can be created with Photogrammetry or through 3D design (as outlined above). The digital asset is then displayed on a generic 3D (in some cases 2D) avatar. Because of costs tied to developing the digital asset, most 3D VTOs still use the Photogrammetry approach, delivering a very approximate and low-quality output. Generative Artificial Intelligence (GenAI) is starting to play a role in this approach, as it can drastically reduce the cost of creating a digital asset. However, due to the early stages of this technology and hallucinations intrinsic in GenAI, the output is far from perfect (e.g. imagine a striped sweater, you will see the stripes intersecting in parts of the sweater) and requires significant manual rework. Google recently launched a GenAI based VTO, and other players include Veesual, Reactive Reality, WEARFITS, amongst many others.
Five different 3D avatars wearing the same digital twin look.

Digital twin of a garment worn by different 3D avatars.

Source: www.veesual.ai

3D VTO with digital twin avatar
3D VTO with digital twin is the holy grail of VTOs. The underlying concept is that of the 3D VTO mentioned above, but displays the image of the product on a digital twin of the consumer, instead of a generic avatar. The digital twin can be created using a picture of the consumer or by virtually recreating a digital avatar that looks like the consumer, with a process similar to that of videogames (think the Sims, or most Role Playing Games). Because of the expected accuracy both from the consumer and the retailer, which is higher if a realistic personalized avatar is being used, this approach is only nascent.
Create-A-Sim mode on The Sims 4.

Create-A-Sim mode on The Sims 4.

Source: www.screenrant.com

What is a Size Advisor?

The other reason a consumer enters a changing room in a physical retail store, is to know what size fits them best. The VTO, as discussed above, helps the consumer understand how a garment looks on them, but not finding the right size. Enter Size advisors. Size advisors aim to support consumers, typically through e-commerce, in identifying the right size to purchase without having to physically try-on a garment. Unlike VTO, players have been trying to solve this problem since the early 2010s. While VTOs aim mostly to increase conversion rates and average order value, Size (and fit) advisors aim to increase conversion rates, average order value, and at the same time reduce size & fit returns, their environmental impact, and the burden on customer service.

Several approaches address this challenge, each differing in the effort required from the retailer, the complexity of inputs requested from the consumer, and the accuracy of the output. Some rely on a pure Machine Learning approach based on sales and returns data to make recommendations. Others aim to create a digital twin of the consumer's measurements in the backend, either through predictive AI, Machine Learning, or computer vision (e.g. asking the consumer to measure themselves using a picture). These methods are analyzed in more detail in the following section.

The dichotomy of the Size Advisor and Virtual Try On

While both Virtual Try-On (VTO) and Size Advisor players are working to solve this need, no comprehensive solution exists on the market at the moment. Why is that?
The reason behind this separation between the two technologies lies in the creation of the digital twin of the product. Creating a digital twin of the consumer is simple from a technological perspective, especially for Size Advisor solutions that recreate a digital twin of the consumer's body measurements. However, once the body is created, it needs to be dressed with a digital twin of the product. This takes us back to some of the challenges discussed above with the digital asset creation process.

Paper patterns making.

Paper pattern making.

Source: www.honeybeefolkschool.ca

Photogrammetry
Assets created with a Photogrammetry approach (discussed in more detail above) do not have physical properties. Hence, they cannot be accurately overlaid on a volumetric model of a person. Furthermore, the cost ($300-$700) of developing an asset creates a high barrier to scalability (think of a retailer with 2500 SKUs per season: 2500*2*$300 = $1.5M per year).
3D design
This highly accurate 3D design process (discussed in more detail above) creates the perfect digital asset for this purpose. However, companies that use software like Clo 3D, Browzwear or Style3D still represent a truly minimal percentage of the fashion market. As fashion companies continue to increase their investment in new technologies —projected by a 2022 McKinsey report to double from 1.6-1.8% of revenues in 2021 to 3-3.5% by 2030— these assets will be more available. However, their adoption in product development is expected to remain limited to a small segment of the industry over the next decade. In the absence of these digital assets from the product development process, external companies re-create them for retailers using the same software, so that they can be used in e-commerce. As discussed above, with a per product cost of $1500-2500$ this is too expensive to be scaled on entire collections (1500$*2500 SKUs * 2 seasons = $7.5M/year).
GenAi
As mentioned before, GenAi could be the true accelerator of the real Virtual Changing Room. If the technology maintains the current pace of improvement (which is a significant “if”), it may become reliable and cheap enough to be combined with a Size Advisor solution that builds the digital twin of the consumer's body.

The benefits of introducing a Size Advisor

The goals brands should be aiming for

Implementing a Size Advisor to help consumers find the right size to purchase can yield substantial benefits to the shopper experience, e-commerce profit & loss, and sustainability of a retailer. Here are some considerations for brands and retailers to evaluate when deciding whether to introduce a Size Advisor, and throughout the selection process:

Reduce costs - Return Rate reduction
This is often the primary motivation for retailers to explore the implementation of a Size Advisor. With fashion return rates averaging 20-40% of sold volumes, and size and fit accounting for 40-60% of total returns, this pain point is easy to identify and quantify.
Reduce costs - Customer Service burden reduction
Customer Service requests are seldom mapped granularly, hence retailers tend to have limited visibility on the drivers behind this cost center. Estimates and client interviews suggest 25-40% of fashion e-commerce customer service calls are related to sizing. At 5-10 minutes per call, this pain point drives both tangible costs, and the intangible cost of customer satisfaction.
Increase sales - Conversion Rate optimization
While this challenge represents the greatest source of loss for online retailers, its nature as an opportunity cost makes it more difficult to quantify unless deliberately monitored. According to studies, 52% of consumers hesitate to complete a purchase when unsure about fit, while a separate 2023 survey indicated that 58% of consumers declare they will never purchase online from a brand they're not familiar with. Convincing even a small portion of these very significant percentages that they are purchasing the right size, can have a significant impact on increasing sales.
Increase sales - Average Order Value improvement
Tied to the above pain point, is the depression of Average Order Value (AoV) due to uncertainty of fit. 49% of consumers interviewed in the same 2023 study, declared they would add 1-2 more items to a shopping session if they could be sure of buying the right size. Giving consumers confidence in the size they're about to purchase, can both increase the number of Units Per Ticket and the mix of more expensive items being purchased.
Improve Sustainability
Whether driven by greenwashing, genuine commitment to the UN's 2050 Sustainable Development Goals, or the need to comply with emerging global regulations (such as France's 2020 Anti-Waste and Circular Economy Law), tackling returns and the excess inventory associated with returns is crucial for achieving these objectives.
Maximize adoption of the solution
In order to maximize the achievement of the above mentioned goals, retailers need to funnel as much traffic as possible to an accurate size recommendation. This will multiply the benefit of an individual KPI (e.g. increase in conversion rate of the users that receive a recommendation) by a higher number of users. Maximizing adoption can be accomplished both by giving the most possible visibility to the tool's Call To Action (CTA), and selecting a tool that has an optimized funnel from open rate to completed recommendation.
Minimize operational effort
Your teams are busy - like never before. According to a 2022 study by the American Psychological Association 60% of employees report negative impacts from work-related stress, notably, 44% experience physical fatigue, marking a 38% increase since 2019. According to 2024 statistics by Happeo 75% of workers have experienced work-related burnout. In this context, when exploring any new product, feature or project, assessing beforehand the impact it will have on day-to-day effort is essential. Understanding which activities will be required to maintain the feature, who these activities will fall on, and distribution of peaks is essential.

Quantifying the benefits of a Size Advisor

Considering the goals outlined above, assessing the impact or benefits of each axis for an individual retailer is a painful yet necessary exercise. Making the exercise additionally challenging is the task of normalizing KPIs, as often different providers will have differing definitions for each indicator. Below is our guide, with our full transparency on the KPIs achievable through Measmerize:

  1. Returns Reduction
While this might seem like a straightforward KPI, different players choose to report this in different ways. The first required distinction is whether the reduction refers to all returns, or only to size and fit related returns. As a second derivative, it's important to clarify if the KPI measures an absolute reduction in returns or the difference in return rates between users who receive recommendations and those who do not. At Measmerize we prefer reporting the difference in return rate between users that follow a recommendation vs. users that do not. The reason for this is that it allows to actually measure the accuracy of the recommendation algorithm, and it is not polluted (or influenced) by the percentage of users that use the tool (adoption rate, a separate KPI). Furthermore, it is important to understand if this KPI differs across categories (e.g. shoes vs. apparel). Measmerize users who follow size recommendations consistently experience a 40% lower return rate for size- and fit-related issues for size- and fit-related issues compared to those who don't, across all categories.
  1. Customer Service calls
Given that Customer Service calls typically go through different IT systems than the e-commerce platform (at least partially), this KPI is seldom measured in a structured manner by Size Advisor providers. At Measmerize, based on qualitative interviews with clients, we've seen a reduction of 20-30% of Customer Service calls requiring size & fit inputs.
  1. Conversion Rate optimization
Just like with returns, defining the perimeter of analysis is very important for this KPI. First of all, we need to define what part of the funnel is measured (Product Listing Page, Product Details Page, Add to Cart, Order/Transaction). Additionally, it is essential to understand if this represents an absolute improvement or a comparison of Conversion Rates (CR) between users that use the Size Advisors and those who don't, or those who use the traditional Size Guide. With a similar logic to returns, we recommend isolating the KPI from adoption of the tool, hence we report the difference in CR between users that receive a recommendation from Measmerize and those that don't, and Measmerize users and users of the traditional Size Guide. Our data shows that users engaging with Measmerize experience a 300-900% higher CR compared to those who don't use the tool. Furthermore, Measmerize users see a 60-80% higher CR compared to users relying on the traditional Size Guide. Our conversion rate is measured from PDP to Add to Cart, which is where the Size Advisor should have the most impact. We also measure and report from PdP to Order.
Measmerize users see a 60-80% higher CR compared to users relying on the traditional Size Guide.

Our conversion rate is measured from PDP to Add to Cart, which is where the Size Advisor should have the most impact. We also measure and report from PdP to Order.

  1. Average Order Value
AoV follows a similar dynamic to returns and CR. It is important to understand if the KPI reports an overall improvement in AoV, or compares the AoV to that of users that don't use the service. To isolate AoV from adoption, at Measmerize we report and see the AoV of Measmerize users being (on average) 13% higher than that of users that don't receive a recommendation from Measmerize.
  1. Sustainability
When reporting the impact on sustainability driven by returns, it is important to understand both the perimeter (shipping, reconditioning, inventory loss…) and the underlying methodology. We will write a separate deep-dive on the methodology, and individual impact is highly influenced by the volumes generated by an individual retailer. Overall, at Measmerize, as of January 2025, we're proud of having avoided 860 tons of CO2 emissions over the past 4 years, equivalent to parking 216 cars/year.
  1. Adoption, or Recommendation Rate
As discussed extensively above, individual KPIs should be multiplied by the number of users that receive a size recommendation. That will allow a brand/retailer to quantify the overall impact. There are four KPIs we recommend breaking down Adoption Rate into:
  1. Open Rate: The percentage of PdP users that click on the CTA of the Size Advisor. As mentioned before, this is mostly impacted by the visibility given to the Size Advisor on the e-commerce page. At Measmerize, where a significant portion of our client-based is luxury players who prefer a minimalist CTA, we see an open rate of 3-6% of PdP users. Beware of providers that report an open rate of a different order of magnitude from the above, at parity of visibility. The user doesn't know which provider is behind the CTA before clicking, so given the same visibility of CTAs the open rate should not vary.
  2. Completion Rate / Funnel: This is the most important KPI when assessing efficacy of a Size Advisor provider, as it is directly impacted by the UX and UI of the solution. At Measmerize we see 75-90% of users who open the tool and interact with anything on the first page completing the process and receiving a size recommendation.
  3. Recommendation Rate: This is a simple multiplication of Open Rate and Completion Rate. As a result of the above KPIs, we see Measmerize clients with a Recommendation Rate (% of PDP users that receive a size recommendation) between 2.5% and 5.5%.
  4. Percentage of Sales based on a recommendation: This is the second most important KPI to monitor in assessing the quality of a size recommendation solution on sales. This is an indirect multiplication of Recommendation Rate and Conversion Rates (and AoV if measured in value, or $, instead of volumes). At Measmerize we manage on average between 15% and 25% of our client's total e-commerce sales.
  1. Return on Investment
The ultimate KPI. This KPI, if measured correctly, combines all of the above KPIs, including pricing of the solution. It should consider impact on Conversion Rates and Average Order Value, multiplied by the Gross Margin of the retailer, and the reduction in costs from Returns (we don't recommend including impact on Customer Service calls as it is unreliable in most cases). At Measmerize, our clients constantly see RoI between 5-10x, with peaks of >20x for larger retailers.
Watch the video 'Do sizing recommendation solutions work?', where Benito Condemi De Felice, Head of Digital Experience and Engagement at Golden Goose, explains why they selected Measmerize as their sizing solution partner over the competition and how they seamlessly integrated Measmerize as a size and fit advisor across multiple categories.
Youtube thumbnail preview of The Golden Goose experience with Measmerize

Do size recommendation solutions work? The Golden Goose experience with Measmerize

Source: www.measmerize.com

How to choose a Size and Fit Advisor provider

Navigating a complex ecosystem

Even if they've truly started taking hold after the Covid-driven fashion e-commerce acceleration, Size Advisors have been around for 15 years. As a result there are many providers out there —at Measmerize we've mapped more than 40 competitors currently. They differ significantly in their approaches to providing a recommendation, using various inputs, data science techniques, and geographical focus. Assortment coverage (e.g. apparel, shoes, kids, jewelry) also varies from player to player.

Launching a solution that provides some sort of size recommendation is surprisingly simple, which does not however mean that the accuracy of the recommendation will be any good. In the following section we aim to provide a framework for retailers to assess the fit (pun intended) of a provider with their business needs.

Breaking down the complexity - the macro models

While there are tens of providers that offer size recommendation services, we feel they can all be clustered into three/four macro models, or approaches, each with different benefits and limitations.

1. Pure Machine Learning (ML)
The Pure ML approach is the one used by the early movers in the industry. TrueFit, Fitanalytics and Secret Sauce Partners (Fit Predictor) are the main players in this space. While their algorithms differ, the underlying approach is the same. They ask the shopper basic questions indirectly tied to their bodies, and cluster consumers based on their responses, under the assumption that people who answer similarly should have similar questions. They then make a recommendation by looking at what size of that product has been purchased and not returned by people from the same cluster.
This approach relies almost exclusively on sales and returns data to build social proof that the size bought by other people in the same cluster should be correct for that consumer. Product measurements bear almost no weight in making the recommendation.

This approach doesn't require product / anthropometric measurements (hence has had a lot of success with multibrand retailers which don't have this information). Furthermore, it has a simple User Experience that maximizes adoption.

Because it heavily relies on sales & returns feedback loop, and not on product measurements, this approach has trouble capturing differences between SKUs, and requires a high volume of sales / SKU to achieve accuracy. Furthermore, it provides more statistical feedback that lacks personalization (e.g. no information on detailed fit of the product on the body).

2. Computer Vision
The idea behind this approach is to digitally measure a consumer using an external device, whether a physical scanner or a mobile phone's camera. This approach is not new either, and has been attempted by 10s of players (Japanese retailer Zozo with Zozosuits, Presize, Styku, Sizer, Friguant, Kleep, 3DLOOK, Mport, Bodi.me, Meepl, Styku, WeWear, 3d-A-Porter, …). It is also the approach Measmerize was initially launched on, before pivoting away from it. The concept is that a consumer's body measurements are extracted (typically from a mobile phone camera, in a controlled setting) and compared to the measurements of the garment.

This approach can be immediately accurate, without requiring a lot of sales & returns data. Furthermore, because it uses both body and garment measurements to make a recommendation it can provide detailed feedback on how a garment fits on the body.

Having to take a picture of yourself, from a specific distance, with a specific outfit (e.g. a hoodie wouldn't allow you to capture the correct outline of the body) requires minutes and effort, making User Experience very complicated. Furthermore, consumers are increasingly worried about privacy. This results in VERY low adoption rates, ultimately minimizing the benefit of the Size Advisor's benefit on both sales and returns. This approach also requires product measurements, making it harder to implement on multibrands.

3. Digital centimeters
The digital centimeters approach aims to make a centimeter (or inches) based recommendation while simplifying the user experience. This method builds an estimate of the consumer's measurements asking simple questions to the consumer, similar to those of the Pure ML model (but uses the answers for a different purpose). The measurements of the consumer are then compared to the measurements of the product (or the anthropometric measurements on which the product was developed). The initial recommendation is then typically fine-tuned over time using Machine Learning, based on sales & returns feedback loop. Measmerize operates with this approach.

This approach is immediately accurate, as it leverages body measurements and compares them to garment measurements. Furthermore, because the User Experience for the consumer is very quick (less than 1 minute) it can also deliver great adoption rates.

The key downside of this approach is that it requires product, or anthropometric, measurements - making it harder for multibrands to implement.

4. Computer Vision Light
Given the low adoption rate of the Computer Vision approach, some players have begun working on a hybrid between the Computer Vision and the Digital Centimeters approach. Essentially, they ask basic questions about the shopper (age, weight, height, gender) and make an initial recommendation based on this. They then ask the shopper to take a picture of themselves, as an optional additional step, to fine-tune the recommendation.

By making an initial recommendation on very basic inputs, they simplify the UX and maximize adoption. They can also provide detailed feedback on how a garment fits on the body.

Taking a picture remains a complex UX step, so most size recommendations rely on basic inputs such as age, weight, height, and gender. This oversimplified approach assumes that all men who are 35 years old, weigh 85 kg, and stand 185 cm tall have identical body types — failing to account for individual differences. These differences are particularly important given that Size Advisor users typically fall outside the "average" body range. Consumers with standard body types are less likely to use a size advisor, whereas those with unique features, like broad shoulders, are more inclined to seek one. As a result, this method often delivers very inaccurate recommendations.

The devil is in the details - what questions should you ask Size Advisor providers

Once a retailer has identified which macro model is best for their business needs and individualities, it becomes even harder to understand the differences between players that use the same approach. Below we provide a guideline of due diligence questions we recommend assessing in order to accurately compare solutions:

1. User Experience
  • What level of customization can be achieved in terms of design and layout, for both the widget (interface) and the call to action?
  • Can the copy — or wording — of the questions be customized to reflect my brand's tone of voice?
  • What are the exact inputs required from the shopper? Which are mandatory and which ones are optional?
  • What is the percentage of users that completes the process, once the solution is opened (see our section on KPIs above for more details)?
  • What is the increase of Conversion Rate and Average Order Value of users that receive a recommendation vs. traditional size guide users, and vs. users that don't use either (see our section on KPIs above for more details)?
User experience designers working on a product flowchart

Teamwork on the product User Experience.

Source: Unsplash

2. Recommendation Accuracy
  • What product information do you require to make a recommendation?
  • How do you check the accuracy of the product information we provide?
  • How do you train your algorithms?
  • What data sets did you use to train the algorithm to make body estimates? How large is it? How geographically diverse? Do you own it?
  • How many recommendations have you made over the past 3 months?
  • What is the geographic breakdown of the recommendations made by the provider? How does it compare to the geographic presence of the retailer's users?
  • How do you account for differences in bodies across geographies and ethnicities?
  • Do you use the same recommendation algorithm across all your clients? If not, how are clients allocated to different algorithms? What are the differences?
  • What is the return rate difference between shoppers who follow the size recommendations and those who don't (see our section on KPIs above for more details)?
Target representing recommendation accuracy

Size recommendation accuracy.

Source: Unsplash

3. Compliance, Security, Architecture
  • Are you compliant with the privacy regulations in my countries of operation (e.g. GDPR, PIPL, California, …)?
  • How many different privacy policies do you have? How are they differentiated by which geography?
  • What is the legal basis for managing consumer data? Do you use consent?
  • What cookies does the solution use? Are these placed based on consent? Where is the consent collected? If collected on the retailers site, how do you ensure it is respected?
  • What is your accessibility level according to USA legislation?
  • What security certificates and external assessments do you have?
  • How often do you do external penetration testing? Can you share the latest result?
  • Where are your servers located? What geographies does each server serve? Where is consumer data stored?
  • Do you transfer EU data out of the EU? China data out of mainland China? If yes, to where, and based on what legal premise?
  • How is data segregated on your databases between different clients?
  • What automated tests do you have to ensure that development doesn't crash the solution during critical times?
Close-up of software for coding

The tech side of product development.

Source: Trevor Vannoy on Unsplash

4. Operational impact
  • What product information do you need to make a recommendation? What information is mandatory and what information is nice-to-have?
  • In what format do I need to provide product data? How often?
  • In what format do I need to provide sales & returns data? How often?
  • Are data flows automated or manual?
  • How does the integration work? Where is the solution integrated into? How much code?
  • Can you break down the process to go-live? What timeline?
  • How long is the onboarding process for the assortment of the first collection? For the following seasons?
  • How is reporting of the solutions performance monitored and assessed? On a dashboard? Business reviews? Does it integrate directly into my existing analytics solution?
  • What KPIs can I see? How can I break them down? To what granularity?
Close-up of software for coding

Dashoboard for data monitoring.

Source: Unsplash

We hope this article was useful for you and your company.

Reach out to our team to discuss how we can support your specific business needs.

Contact us